Qualcomm moves to loosen Nvidia’s AI grip

Qualcomm confirmed it is acquiring Modular in an all-stock deal valued at approximately $3.92 billion, based on Qualcomm’s Tuesday closing price. Modular’s Mojo language and MAX inference engine are designed to let developers run AI models across chips from Nvidia, AMD, Intel, Qualcomm, Apple Silicon and CPU platforms without hardware-specific rewrites.

The deal anchors a broader push to challenge Nvidia’s hold on AI infrastructure, which rests on GPU hardware and the CUDA software ecosystem. Approximately 4 million developers work within CUDA, and code written for that platform typically requires significant rewrites to run on rival silicon, keeping many customers tied to Nvidia even when alternatives may fit specific workloads.

Qualcomm is also in reported talks to acquire Tenstorrent at a valuation of between $8 billion and $10 billion, though no deal has been announced. Tenstorrent’s RISC-V-based Blackhole chip uses the company’s Tensix architecture for inference workloads, while Modular would provide a compiler layer to reduce the software barriers that have slowed adoption of non-Nvidia AI chips.